Robot-assisted bronchoscopy for pulmonary lesion diagnosis: results from the initial multicenter experience

The increasing need to efficiently and safely sample lung lesions has led to the development of virtual bronchoscopy (VB), radial endobronchial ultrasound (r-EBUS), electromagnetic navigation (EMN), fluoroscopy-based navigation, bronchoscopic trans-parenchymal nodule access (BTPNA), ultrathin bronchoscopes, and cone-beam computed tomography (CBCT) guided-bronchoscopy. The diagnostic yield for lung lesions using these modern bronchoscopic techniques continues to be suboptimal and is 40–70% [1‐7].

The now commercially available robotic endoscopic system (RES; Auris Health, Inc., Redwood City, CA) has been recently FDA-approved for sampling lung lesions. In cadaveric models, robot-assisted bronchoscopy (RAB) was shown to have improved reach in the periphery of the lung in all segments when compared with 4.2 mm OD conventional thin bronchoscopes (by average generation count: 8.7 vs 5.6) [8]. This may be explained by 1) the improved structural support provided by the outer sheath, that is usually locked in the target segment (usually 3rd-4th generation) before advancing the scope, and 2) the improved ability to make subtle turns due to 4-way steering and a distal section capable of achieving articulation in pitch and/ or yaw. RAB also allows direct visualization of peripheral airways and of the biopsy tools as they are advanced outside the working channel, thereby enabling the operator to better steer the tools towards the target. The scope can be locked in position and the instruments advanced through the working channel without exertion of torque onto the bronchoscope, minimizing airway distortion. This, along with the better column strength and telescoping design, could enable higher diagnostic yields of peripheral lesions.

To date, only a small feasibility study that enrolled 15 patients using the RES was performed and showed no pneumothoraces or significant bleeding [9]. The aim of our study is to report on the initial post-marketing feasibility, safety and diagnostic yield of this technology implemented at four US-based university and community centers.

During the study period at the four study centers, 167 lesions were biopsied in 165 patients. Two lesions were biopsied in the same procedure in two cases. The average follow-up was 185 ± 55 days.

This is the first study after the market release of robotic bronchoscopy in March 2018. Our study’s patients’ characteristics and average lesion size are similar to published EMN studies [12‐14]. Navigation success was achieved in 88.6% with 69.1–77% overall diagnostic yield (conservative and maximum estimate). There were 13 cases in which pathology showed inflammation for which follow-up was not available. These cases were considered as non-diagnostic to provide conservative estimates of navigation success and yield based on various lesion characteristics. We provide an overall diagnostic yield range as some of these lesions could have resolved with time and have been diagnostic, if we had long-term follow-up on them [12]. We believe our definition of navigation success is meaningful for clinicians as it consists of the presence of diagnostic material on final pathology, or r-EBUS image confirmation. We did not rely only on the target image generated by the EMN software, which could be prone to multiple errors [15]. However, our definition may be subject to overestimation as it is possible that atelectasis or alveolar filling may have resulted in false positive r-EBUS images. In the absence of CBCT and confirmation of tool-in-target, it is difficult to precisely define true navigation success. In this series, diagnostic yield did not depend on lesion lobar location, centrality or size. The diagnostic yield did not depend on lesion density as well; however, we only had 17 lesions that were pure ground-glass density. The yield was better when a concentric r-EBUS view was obtained compared to an eccentric view (81.5 and 71.7%). The high yield even when an eccentric view was obtained compares favorably to previously reported rates of 48% in such cases [16]. This is likely because RAB allows stability, visualization of the point of contact of the radial probe with the airway wall and enables directional targeting of instruments [17] (Fig. 1). Biopsies were obtained in 84.6% (22/26) of cases in which r-EBUS confirmation was not obtained. The majority of these lesions (73%) were solid. The decision to biopsy was at the discretion of the operator when it was believed that the robotic scope was in the target’s proximity but in an adjacent airway based on the EM-generated target view. Our yield of 54% in the absence of bronchus sign is higher than previously reported rates of 31–44% [18, 19] but is lower than a recently published multi-center EMN study [12].

Procedure-related complications are comparable to those from other guided bronchoscopy studies, with a 3.6% pneumothorax and 2.4% bleeding rate [2, 4, 12]. While bleeding during RAB can be managed with injection of cold saline through the working channel of the robotic scope, in cases of suspected significant bleed, we have decided to disconnect the robotic scope and introduce a therapeutic flexible bronchoscope (2.8 mm working channel) to evaluate and potentially control the hemorrhage. While switching of scopes runs the risk of losing a wedged position and anatomical orientation in situations with major bleeding, we have not yet encountered such bleeding in any of our cases, likely because the robotic scope locked in a wedged position in a small peripheral airway allows containment and clotting of biopsy-related bleeding. All our cases of bleeding were controlled with just cold saline. Given that we injected cold saline prior to suctioning out the blood, we were not able to accurately quantify the volume of blood loss in each case by just looking at the return in the suction canister. There were no cases of respiratory failure or death in our study.

Our case series has several limitations. Our average follow-up period of 6 months is not sufficient to determine the true diagnostic yield and in fact large studies usually report outcomes at 12 months [2, 12]. However, as highlighted above, for the purposes of all calculations, cases that required follow-up that had not been done yet were considered as non-diagnostic. Needles and forceps were used in 100 and 96% of cases, respectively. The order in which they were used was per operators’ discretion and not captured in this analysis. In addition, the tool-specific diagnostic yield was not possible to analyze as the touch prep from forceps biopsies are reported under the cytology section and not labeled as distinct from the needle specimens. We believe that prospective studies should address the independent diagnostic yield for needles, brushes and forceps biopsies. Amongst the 71 patients diagnosed with lung cancer, adequacy of tissue obtained for genetic testing was not reported consistently in the four centers, as the practice of molecular testing for early stage lung cancer remains institution-dependent. Our definition of significant bleeding is unconventional. Unfortunately, with the robotic scope positioned in distal airway, it is very difficult to accurately assess the severity of bleeding. Future prospective studies may be able to better elucidate the true bleeding rate with RAB, without withdrawing the robotic scope. Bronchoscopy room set up, navigation and procedure times were not prospectively recorded in all cases; in the cases wherein this was documented in the medical records, the mean navigation and procedure times were 17.8 and 58.6 min, respectively. Our procedure time only reflects the robotic bronchoscopy portion of a procedure and does not include the time it took to stage the mediastinum with EBUS. Based on these data, duration of navigation and biopsy seem to be similar to other EMN-guided procedures [12].

Despite our results not demonstrating superior diagnostic yield compared to some other recent EMN-guided bronchoscopy studies [12], we believe that in the future, robotic bronchoscopy platforms may eventually enable operators to more precisely navigate to the periphery of the lung and potentially allow for bronchoscopic therapeutic ablation of malignant lesions. As of now, in our opinion, when available, robotic bronchoscopy should be offered to all patients with suspicious peripheral lung lesions that also require 1) concurrent guidelines-recommended EBUS-TBNA lymph node staging for CT-PET normal mediastinum or prior to SBRT; or 2) preoperative tissue diagnosis based on questionable operability, patient or surgeon’s preference.

The results of this analysis suggest that in patients with lung lesions requiring biopsy, post-marketing RAB implementation in community and academic centers is safe with initial diagnostic yield and complication rates similar to existing technologies. Long-term follow-up is required to better establish the true diagnostic yield and delineate the factors affecting it. Comparative trials with existing guided bronchoscopy platforms are warranted for determining cost-effectiveness of this technology in diagnosing lung nodules.